A paper by a former postdoc at MD Anderson Cancer Center who “admitted to knowingly and intentionally falsifying” a figure has been retracted.
In August, the Office of Research Integrity announced that it had sanctioned Jun Fu for faking data in a study of the results of a mouse study of NVP-HSP990, a Novartis compound designed to fight brain tumors. Here’s the notice for the study in question, published in Cancer Research:
The authors wish to retract the article entitled “Novel HSP90 InhibitorNVP-HSP990 Targets Cell-Cycle Regulators to Ablate Olig 2-Positive Glioma Tumor–Initiating Cells,” which was published in the May 15, 2013 (73:3062) issue of Cancer Research (1).
A University of Texas MD Anderson Cancer Center internal investigation has found that Dr. Jun Fu committed research misconduct as defined by 42 CFR Part 93 and MD Anderson’s Research Misconduct Policy. Specifically, that Dr. Jun Fu falsified the data appearing in Figure 8B by modifying the death dates of mouse specimens to produce a fabricated survivability curve for GSC-11. Dr. Jun Fu has accepted full responsibility for this breach of research integrity. This constitutes a violation of MD Anderson policy concerning the ethical conduct of research, as well as a significant departure from the research practices commonly accepted by MD Anderson research community. Because of the severity of this breach of research integrity, the authors believe that retraction of the article represents the most appropriate course of action to ensure the accuracy of the scientific record.
The internal panel further concluded that none of the other co-authors of this article knew, or had reason to suspect that the data presented in the article had been manipulated by Dr. Jun Fu.
All co-authors have agreed to this retraction.
Hat tip: Rolf Degen
Will Jun Fu reimburse the money he received (part or all) from the Caroline Ross Endowment Fellowship so that it can be rerouted to scientists with a better sense of research ethics?
“The internal panel further concluded that none of the other co-authors of this article knew, or had reason to suspect that the data presented in the article had been manipulated by Dr. Jun Fu.”
When it comes to death plots, who’s going to know exactly how long the animals lived and died? Was Jun Fu the “animal guy” in the lab and perhaps the only one with access to the animal facilities (resulting in less questions about the death curve)? Or are vet techs most likely to check on the animals not allowed to speak up at lab meetings when they see KM plots inconsistent with their own notes on when animals died? I assume the former is the case, since maintaining vet techs in a lab is pricier than using apprentice (graduate and post-doctoral) labor.
Novartis has a stake in this too. Pharma outsources to universities because it’s cheaper than doing things in-house. Perhaps the pharma approach is to simply send the library of compounds to academia with an NDA; and every compound that is “a winner” is checked in-house before the academic research lab is given the greenlight to publish. All losing compounds are likely to be losers, but the winning compounds are more likely to have been juiced to make the researcher look good.
Too cynical?
I don’t think it is reasonable to expect co-authors to look up lab records of animal survival and animal techs probably have 20 different experiments on the go for different scientists to care much about this N=10 group. I suppose he felt if the compound had no effect the entire raison-d’etre of his paper was gone. One wonders if all the molecular findings that supposedly underpinned this pharmaceutical effectiveness are genuine or not
As far as it is goes, it looks like the B-actin band from figure 5 GSC-11 Panel B has been reused in figure 7 GSC-11 Panel D
Take a look at Figure 8 (I) and (II)
http://cancerres.aacrjournals.org/content/73/10/3062/F8.large.jpg
Now at 5 mice per cage, that is 2×2 cages, where one had drug and one vehicle. Assume that the drug is identical to saline and the death curve is normal centered 55 days and 15 days SD.
The researcher arrives at the vivarium at 8:00 and records living/unsymptomatic numbers in control and drug treated.
Every mouse that dies, or is at (or near) and ethical endpoint, in the control group is recorded as control, every death in the treatment group is recorded as control, and a mouse from the mirror cage is swapped.
This is why the injector should not know which group is which, you give them vial’s marked T(top) and B(bottom) or L(Left) and R(Right). That way you have real randomization as the randomizer doesn’t know if you will pick L/F or T/B.
The way to do it properly is daily weight of each mouse, in each cage, and have the weights recorded in a Legal Lab book, with each cage recorded and photographed (iPhone), daily.
The necrosis in brain tumors never looks like A(I) either. Necrotic tissue will not generally fix or stick to the slide and you get voids and adhesion artifacts.
Doc, good idea of course. And the recording of weights and death dates etc. should be done by someone not otherwise directly involved in the investigation. Impartial vet techs whos jobs don’t depend on any outcome. 4 eyes principle everywhere. etc.pp.
However, we all no thats not how it works. As Deidentified noted, quite often the one researcher (post-doc or PhD student or grad student) is doing all the research work alone in the animal facility and the vet techs are only doing basic animal care.
But even if vet techs record the important data points in a bullet proof way, chances are high noone in the co-author team checks the raw data that the first author puts into these nifty curves and figures.
Almost all intentional fraud could be detected within a research group, if PIs were only willing to implement basic checks and quality management.
Apparently, Jun Fu did not falsify the original notes. Otherwise, it would have been impossible to prove that he modified death dates.
We, and the researchers at MDA, are charged for animal inspections, weighing and for data recording, even some PI’s are cheaper.
We have a blind system where the vials are given to the injector as A, B and C, or something similar. I make up the drugs and internal controls and label them. Then I record on the inside of the back cover of my lab book the contents, I sign and date, and then have a counter signature by someone outside of my group.
I have the injector, my technician, mix the cages the day before, and then I ask her to pick a layout for the injections. She knows not which is which. She inspects 3-4 mornings during the week and we alternate during the weekends.